1 GENERAL INSTRUCTIONS
Due to production equipment, production environ-
ment, operators and other factors, presswork will ap-
pear misregister, misting, stain, leakage and many
other defects in the high speed printing[1]. All the
difference between standard template and printing
image is printing defects. The online print defect de-
tection system based on machine vision can meet the
real-time requirements of printing quality detection
in industrial production. Many researchers do a lot of
experiments and research on printing defects detec-
tion, and their basic thought was the same. Print de-
fects performance in the image is the difference of
the gray value between the defect image and stand-
ard image. Compare the gray value of defect image
with standard image, and determine the difference is
beyond the scope of preset threshold, so that decide
the presswork with defects or not[2].
In this paper, by using wavelet packet algorithm to
isolate defect area, then take the advantage of
regional analysis method to extract defect features
effectively and quickly, thereby judging defect types.
2 THE BASIC THEORY AND PRINCIPLE OF
WAVELET PACKET TRANSFORM
Due to the orthogonal wavelet transform only further
decomposition to the low frequency part of signal,
but don’t further decomposition to the high
frequency part which is the details, so the wavelet
transform can effectively represent a large family of
low frequency information as the main component
signal, but it can’t well decomposition and
represent the signal that contains a large number of
details(small edge or texture),such as non-stationary
machinery vibration signal, remote sensing images,
seismic signal, biomedical signal. In contrast, the
wavelet packet transform can provide a finer
decomposition for the high frequency part, and the
decomposition without redundancy, no omissions, so
it can be better time-frequency localization analysis
to the signal that contains a large number of medium
and high frequency information.
2.1 The definition of wavelet packet
Assume
zj
j
V
is the multi-resolution analysis of
)(
2
RL [3],
1j
W
is the orthogonal complement space
of
1j
V
in
j
V
, )(
0
tu and
)(
1
tu
are the correspond-
ing scaling function and wavelet function, write for:
zk
k
zk
k
ktugtu
ktuhtu
)2(2)(
)2(2)(
01
00
(1)
Among them:
zk
k
h
,
zk
k
g
,
k
k
k
hg
1
)1(
,
then:
A set of functions define by
zk
nkn
zk
nk
ktugtu
ktuhtu
)2(2)(
)2(2)(
12
n2
(2)
Is
)(tu
n
(
2,1,0n
),this is the wavelet
packet determined by orthogonal scaling function
)(
0
tu [4].
Kind of image defect detection algorithm based on wavelet packet
transform and blob analysis
LiXia.AO & XingBin.ZHOU
Compute Center of Nanchang University, Nanchang, China
ABSTRACT: By using wavelet packet, the image can be decomposed both in the low frequency part and high
frequency part. Wavelet packet analysis technology have been used in image processing as a new time-
frequency analysis tool. In this paper, by using wavelet packet to decompose image and proper threshold to
process the coefficients of the wavelet packet tree, then reconstruct the defect image. Using blob analysis
method to extract the defect object and determine its type.
KEYWORD: wavelet packet; Blob Analysis; printing defects